Learning FEA and Applications with Concrete

[craigmcg]

Hi all,

I've spent bits of free time over the past year or so trying to learn what I can about FEA and practical application in structural analysis for civil structures. I haven't spent a great deal of time on in-depth theory just yet, but I plan on taking some night courses once this d*mn economy turns around. Predictably the deeper I go, the more questions I have. I don't have a lot of resources at the moment, so I was hoping that I could pick the brains of you folks. I'm hoping that if I can get some of the basic application aspects figured out, it will help me when I get back into the classroom.

I deal mainly with concrete in my designs. Occasionally steel, aluminum, masonry and vinyl, but concrete is my primary area. As a case example, imagine a 8'x40' elevated concrete platform supported on piles. My questions are these:

1) Say the platform is 1' thick (nevermind subsequent issues such as punching shear etc). I would like to model the slab in FE while the piles are modeled as frame elements. Given this thickness, would it be preferable to model the slab as a plate or shell? I understand that one has one more DOF than the other, but how do you determine exactly which in a particular application? Any rules of thumb?

2) Say the platform is increased to 5' thick, would this require the model to be revised to solid elements? Does reinforcement and concrete have to be modeled seperately in FEA with solids?

3) Say the piles are widely spaced along the 40' length and a torsional moment was induced into the platform. Is it even possible to design plate or shell elements for torsion? Or would this require solids as well? Secondly, how do you interpret the torsional force in ksi for concrete design with ACI where a kip-in force result is required (Wood-Armer doesn't apply here)?

4) Are reinforced concrete FE solid models even valid considering that concrete has no appreciable tensile capacity? Unless this was accounted for in the model of course.

5) At the connection of the piles into the cap, is it correct to assume a master/slave node about the center of the pile to it's extremities? Otherwise I find that slab stresses get unrealistically high and localized.

Hope this made sense. Feel free to disagree anything I've said if its incorrect. Thanks in advance.

 

[IDS]

1) You need to look at how the actual structure is free to move, and make sure that the element you use can move in the same way. A floor slab has out of plane deformations, so you will need a shell element (sometimes called a plate/shell element) if you are doing a 3D analysis. 2D plate elements only consider the stresses in the plane of the element, and may be plane stress (e.g. for a shear wall) or plane strain for a 2D section of a long structure like a tunnel.

2) You need to think about the information you actually need for the design. In this case you might be better doing a strut-tie analysis in the support regions and not worrying about FEA. You might still want to model the overal structure, in which case the degree of detail you went to would depend on the importance of this element to the overall behaviour. If it is significant you might want to make increasing refinements of the model, until two successive models gave substantially the same results.

It would be unusual to model reinforcement and concrete separately in a 3D analysis for design work. It gets very complicated because you need to handle cracking, slip of the reinforcement relative to the steel, and such like.

3) It's difficult to anser this without more details, but again I would recommend reveiwing what you actually need for the design, and whether you really need to look at torsion (and if you do, will a simplified analysis be adequate?).

4) If the stiffness of the members you are using is significant you certainly need to consider whether it will be cracked or uncracked. It may be adequate to do an uncracked analysis, and a second one with fully cracked properties, or you may need to model the actual behaviour more carefully. Using shell elements you will probably be able to apply a moment curvature table to the elements. Doing this with brick elements is more difficult, and again is not something normally done for design purposes.

5) You will often need a rigid (or stiff) connection at the intersections of linear members. You can either use a beam element with suitably high stiffness, or use a rigid connector element. A master/slave link is probably not what you want because it will make the connected nodes move equally in space, rather than allowing a rotation. Again you need to look at how the model will behave, and does this reflect the behaviour of the actual structure.

Hope that helps

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/